movenc: Use the actual converted sample for RTP hinting

If an annex b bitstream is muxed into mov, the actual written
sample is reformatted to mp4 syntax before writing.

Currently, the RTP hints that copy data from the normal video
track, where the payload data might be offset compared to the
original sample that the RTP hinting used (when 3 byte
annex b startcodes have been converted into 4 byte mp4 format
startcodes).

Signed-off-by: Martin Storsjö <martin@martin.st>

libavformat/movenc.c

@@ -2002,6 +2002,7 @@ int ff_mov_write_packet(AVFormatContext *s, AVPacket *pkt)
     AVCodecContext *enc = trk->enc;
     unsigned int samplesInChunk = 0;
     int size= pkt->size;
+    uint8_t *reformatted_data = NULL;
 
     if (!s->pb->seekable) return 0; /* Can't handle that */
     if (!size) return 0; /* Discard 0 sized packets */
@@ -2035,7 +2036,13 @@ int ff_mov_write_packet(AVFormatContext *s, AVPacket *pkt)
     if (enc->codec_id == CODEC_ID_H264 && trk->vosLen > 0 && *(uint8_t *)trk->vosData != 1) {
         /* from x264 or from bytestream h264 */
         /* nal reformating needed */
+        if (trk->hint_track >= 0 && trk->hint_track < mov->nb_streams) {
+            ff_avc_parse_nal_units_buf(pkt->data, &reformatted_data,
+                                       &size);
+            avio_write(pb, reformatted_data, size);
+        } else {
             size = ff_avc_parse_nal_units(pb, pkt->data, pkt->size);
+        }
     } else {
         avio_write(pb, pkt->data, size);
     }
@@ -2090,7 +2097,9 @@ int ff_mov_write_packet(AVFormatContext *s, AVPacket *pkt)
     avio_flush(pb);
 
     if (trk->hint_track >= 0 && trk->hint_track < mov->nb_streams)
-        ff_mov_add_hinted_packet(s, pkt, trk->hint_track, trk->entry);
+        ff_mov_add_hinted_packet(s, pkt, trk->hint_track, trk->entry,
+                                 reformatted_data, size);
+    av_free(reformatted_data);
     return 0;
 }
 

libavformat/movenc.h

@@ -120,7 +120,8 @@ int ff_mov_write_packet(AVFormatContext *s, AVPacket *pkt);
 
 int ff_mov_init_hinting(AVFormatContext *s, int index, int src_index);
 int ff_mov_add_hinted_packet(AVFormatContext *s, AVPacket *pkt,
-                             int track_index, int sample);
+                             int track_index, int sample,
+                             uint8_t *sample_data, int sample_size);
 void ff_mov_close_hinting(MOVTrack *track);
 
 #endif /* AVFORMAT_MOVENC_H */

libavformat/movenchint.c

@@ -95,11 +95,12 @@ static void sample_queue_free(HintSampleQueue *queue)
  * not copied. sample_queue_retain should be called before pkt->data
  * is reused/freed.
  */
-static void sample_queue_push(HintSampleQueue *queue, AVPacket *pkt, int sample)
+static void sample_queue_push(HintSampleQueue *queue, uint8_t *data, int size,
+                              int sample)
 {
     /* No need to keep track of smaller samples, since describing them
      * with immediates is more efficient. */
-    if (pkt->size <= 14)
+    if (size <= 14)
         return;
     if (!queue->samples || queue->len >= queue->size) {
         HintSample* samples;
@@ -109,8 +110,8 @@ static void sample_queue_push(HintSampleQueue *queue, AVPacket *pkt, int sample)
             return;
         queue->samples = samples;
     }
-    queue->samples[queue->len].data = pkt->data;
-    queue->samples[queue->len].size = pkt->size;
+    queue->samples[queue->len].data = data;
+    queue->samples[queue->len].size = size;
     queue->samples[queue->len].sample_number = sample;
     queue->samples[queue->len].offset = 0;
     queue->samples[queue->len].own_data = 0;
@@ -386,7 +387,8 @@ static int write_hint_packets(AVIOContext *out, const uint8_t *data,
 }
 
 int ff_mov_add_hinted_packet(AVFormatContext *s, AVPacket *pkt,
-                             int track_index, int sample)
+                             int track_index, int sample,
+                             uint8_t *sample_data, int sample_size)
 {
     MOVMuxContext *mov = s->priv_data;
     MOVTrack *trk = &mov->tracks[track_index];
@@ -402,7 +404,10 @@ int ff_mov_add_hinted_packet(AVFormatContext *s, AVPacket *pkt,
     if (!rtp_ctx->pb)
         return AVERROR(ENOMEM);
 
-    sample_queue_push(&trk->sample_queue, pkt, sample);
+    if (sample_data)
+        sample_queue_push(&trk->sample_queue, sample_data, sample_size, sample);
+    else
+        sample_queue_push(&trk->sample_queue, pkt->data, pkt->size, sample);
 
     /* Feed the packet to the RTP muxer */
     ff_write_chained(rtp_ctx, 0, pkt, s);